A dependable website backup strategy combines the right backup types, a retention schedule that matches your recovery needs, and regular restore tests. This checklist explains how often to back up a website, what to keep, where to store it, and how to confirm that a backup can actually bring your site back online.
Overview
A backup is useful only if it is recent, complete, accessible, and restorable. Many hosting plans create automatic copies, but those copies may have limited retention, remain on the same infrastructure as the live site, or cover only selected data. Treat hosting backups as one layer of protection rather than your entire disaster-recovery plan.
Start by documenting what must be recovered: website files, databases, uploaded media, configuration files, environment variables, email-related settings, DNS records, and any external services connected to the site. Then define two targets:
- Recovery point objective (RPO): the maximum amount of recent data you are willing to lose. If your site changes several times per day, a daily backup may not be sufficient.
- Recovery time objective (RTO): the time you can reasonably allow before the site needs to be operational again. A small brochure site and a busy store may require very different procedures.
Keep at least one copy separate from the hosting account. A compromised hosting account, accidental deletion, failed migration, or ransomware event can affect both the live site and backups stored beside it.
Backup types compared
| Backup type | What it captures | Best use | Limit to check |
|---|---|---|---|
| Hosting snapshot | A point-in-time image of a server, account, or storage volume | Fast rollback after a configuration or deployment problem | Whether it includes databases, external storage, and enough historical versions |
| Full-file backup | Application files, themes, plugins, uploads, and configuration files | Rebuilding the file layer on the same or a new host | Whether sensitive files are encrypted and whether permissions are preserved |
| Database export | Posts, pages, orders, users, settings, and other structured content | Recovering changing content or moving a CMS | Database engine, version compatibility, character set, and export completeness |
| Off-site copy | A separate copy stored outside the primary hosting environment | Protection from account loss, infrastructure failure, or destructive incidents | Access credentials, encryption, retention, and restore speed |
Checklist by scenario
Low-change brochure or information site
For a site that changes occasionally, schedule a full backup at least before every major update and on a recurring cadence that matches the value of the content. A practical baseline is a weekly backup, with an additional copy before plugin, CMS, theme, code, or hosting changes.
- Back up website files and the database together.
- Save a copy outside the hosting account.
- Record the CMS version, PHP or runtime version, and important extensions.
- Test a restore after the initial setup and periodically thereafter.
Frequently updated publishing site
News, documentation, membership, and community sites can lose significant work between weekly backups. Use more frequent database backups, such as daily or several times per day, while retaining less frequently changing files separately. Confirm that uploads are included; a database-only export will not restore images, downloads, or media libraries.
- Set the database schedule according to publishing volume.
- Back up uploads and application files on a separate schedule if appropriate.
- Keep enough versions to recover from an error discovered days later.
- Check that scheduled jobs complete and alert someone when they fail. See the cron jobs guide for reliable scheduling and monitoring practices.
Online store or transaction-based site
For a store, losing orders, customer records, inventory changes, or payment-related configuration can have operational consequences. Backups should be frequent enough to meet the agreed RPO, and restoration should be tested in a non-production environment. Also identify which records live in external payment, shipping, email, or analytics systems; a website backup may not contain those records.
- Use frequent database protection for orders and inventory.
- Coordinate backups with maintenance windows and deployments.
- Confirm that restored checkout, webhooks, scheduled tasks, and email notifications do not accidentally operate against live services.
- Document a read-only or maintenance-mode procedure for emergency recovery.
Website migration or major change
Before changing hosts, DNS, database versions, plugins, themes, or application code, create a known-good backup and label it clearly with the date and environment. Do not overwrite that copy during the migration. Restore it to a staging environment first, compare the result with the live site, and keep the original hosting account available until validation is complete. The website migration checklist can help coordinate this work.
3-2-1 backup checklist
Use the 3-2-1 model as a planning test:
- 3: maintain three copies of important data, including the live version.
- 2: use at least two different storage locations or media types.
- 1: keep at least one copy off-site and separate from the primary hosting account.
For higher-risk environments, consider making one backup difficult to alter or delete through access controls, immutability features, or offline storage. Choose controls that your team can administer and restore from under pressure.
What to double-check
Completeness
List every component required to run the website. For a typical CMS, that may include the database, core files, themes, plugins, uploads, custom code, web-server rules, certificates, environment variables, and deployment configuration. DNS records and domain registration are usually managed separately, so document them independently. A DNS change verification workflow can help confirm that records still point to the intended environment.
Retention and naming
A retention policy should state how many daily, weekly, and monthly versions are kept, when older copies are deleted, and who can change the policy. Use names that include the site, environment, date, and backup type, such as example-prod-db-2026-11-05. Keep a simple worksheet with these fields:
- Site and environment:
- Backup type and included components:
- Schedule and time zone:
- Retention period:
- Primary and off-site locations:
- Encryption and access owner:
- Last successful restore test:
- Expected RPO and RTO:
Security and access
Restrict backup access to the people and systems that need it. Use separate credentials where possible, protect storage with encryption, and avoid placing backup archives in a publicly accessible web directory. If backups contain personal information, customer data, or credentials, include them in your broader data-protection review. Rotate access when team members, hosting providers, or deployment systems change.
Restore testing
A successful backup job is not proof of a usable backup. At least periodically, restore a copy to staging or an isolated environment:
- Select a backup by date and verify its checksum or archive integrity if your tools provide one.
- Prepare a clean environment with compatible runtime, database, and web-server versions.
- Restore the files and database using documented steps.
- Update configuration without exposing staging to public traffic or live payment services.
- Test the homepage, representative content, login, forms, media, search, scheduled tasks, and critical integrations.
- Record the elapsed restore time, errors, missing components, and corrective actions.
- Delete temporary credentials and restrict or remove the restored environment when testing is complete.
For WordPress or another CMS, test both the database and the media library. For custom applications, include dependency installation, build steps, secrets management, and deployment configuration in the recovery documentation.
Common mistakes
- Relying only on hosting snapshots: snapshots may not meet your retention or off-site requirements. Confirm exactly what the host captures and how restoration works.
- Backing up files but not the database: the site may load while current content, accounts, or orders are missing.
- Storing every copy in one account: an account lockout or destructive incident can remove the live site and its backups together.
- Never testing a restore: incomplete archives, incompatible versions, missing credentials, or undocumented steps may surface only during an emergency.
- Keeping backups indefinitely without a plan: excessive retention increases storage, access, and privacy concerns. Define what is retained and why.
- Ignoring DNS, domains, and certificates: a restored server is not enough if the domain points elsewhere or HTTPS configuration is missing. Review domain renewal settings and SSL certificate procedures separately.
- Overlooking scheduled tasks and integrations: cron jobs, webhooks, email delivery, payment services, and external APIs may need reconfiguration after a restore.
When to revisit
Review this website backup strategy before seasonal planning cycles, major releases, hosting migrations, CMS upgrades, infrastructure changes, or periods of increased transaction volume. Revisit it whenever the workflow or tools change, including a new backup provider, deployment pipeline, database engine, storage location, or access-control model.
At least annually, run a disaster-recovery review with a real restore test. Compare the measured restore time with the target RTO, confirm that the newest backup meets the target RPO, remove obsolete credentials, and update the inventory of files, databases, DNS records, certificates, and external services. If the website has grown, adjust backup frequency and retention rather than assuming the original schedule still fits.
Action plan: document the site components today, enable a scheduled backup, create an off-site copy, label the retention rules, and perform one controlled restore. Put the next restore test on the calendar before you finish. A backup strategy becomes dependable through repeated verification, not through configuration alone.